{
  "name": "colloquial_tool_router_repair_loop",
  "status": "failed_acceptance_thresholds",
  "acceptance_thresholds": {
    "structured_tool_eval_overall": ">=95%",
    "generic_prompt_correct": ">=80% including weather/browser and terminal",
    "text_tool_leaks": "0",
    "no_tool_false_positive": "<=5%",
    "tool_result_finalization": ">=90%"
  },
  "main_failure": "terminal and repo-search colloquial prompts still route to browser_navigate or bash under structured OpenAI tool_calls",
  "eval_summaries": {
    "iter02_pretrain_colloquial_openai_parser_disabled": {
      "passed": 16,
      "total": 20,
      "overall_rate": 0.8,
      "tool_cases": 13,
      "structured_tool_cases_passed": 9,
      "structured_tool_rate": 0.6923,
      "no_tool_cases": 4,
      "no_tool_cases_passed": 4,
      "false_positive_tool_rate": 0.0,
      "finalization_cases": 3,
      "finalization_passed": 3,
      "finalization_rate": 1.0,
      "finish_reason_tool_calls": 13,
      "content_text_tool_leaks": 0,
      "category_metrics": {
        "browser": {
          "passed": 4,
          "total": 4,
          "rate": 1.0
        },
        "calculator": {
          "passed": 2,
          "total": 2,
          "rate": 1.0
        },
        "file": {
          "passed": 2,
          "total": 3,
          "rate": 0.6667
        },
        "finalization": {
          "passed": 3,
          "total": 3,
          "rate": 1.0
        },
        "no_tool": {
          "passed": 4,
          "total": 4,
          "rate": 1.0
        },
        "terminal": {
          "passed": 1,
          "total": 4,
          "rate": 0.25
        }
      },
      "failures": {
        "generic_terminal_echo": "wrong_tool:browser_navigate",
        "generic_terminal_pwd": "wrong_tool:browser_navigate",
        "generic_terminal_ls": "wrong_tool:bash",
        "generic_search_repo": "wrong_tool:browser_navigate"
      }
    },
    "iter02_colloquial_openai_parser_disabled": {
      "passed": 16,
      "total": 20,
      "overall_rate": 0.8,
      "tool_cases": 13,
      "structured_tool_cases_passed": 9,
      "structured_tool_rate": 0.6923,
      "no_tool_cases": 4,
      "no_tool_cases_passed": 4,
      "false_positive_tool_rate": 0.0,
      "finalization_cases": 3,
      "finalization_passed": 3,
      "finalization_rate": 1.0,
      "finish_reason_tool_calls": 13,
      "content_text_tool_leaks": 0,
      "category_metrics": {
        "browser": {
          "passed": 4,
          "total": 4,
          "rate": 1.0
        },
        "calculator": {
          "passed": 2,
          "total": 2,
          "rate": 1.0
        },
        "file": {
          "passed": 2,
          "total": 3,
          "rate": 0.6667
        },
        "finalization": {
          "passed": 3,
          "total": 3,
          "rate": 1.0
        },
        "no_tool": {
          "passed": 4,
          "total": 4,
          "rate": 1.0
        },
        "terminal": {
          "passed": 1,
          "total": 4,
          "rate": 0.25
        }
      },
      "failures": {
        "generic_terminal_echo": "wrong_tool:browser_navigate",
        "generic_terminal_pwd": "wrong_tool:browser_navigate",
        "generic_terminal_ls": "wrong_tool:bash",
        "generic_search_repo": "wrong_tool:browser_navigate"
      }
    },
    "iter03_colloquial_openai_parser_disabled": {
      "passed": 16,
      "total": 20,
      "overall_rate": 0.8,
      "tool_cases": 13,
      "structured_tool_cases_passed": 9,
      "structured_tool_rate": 0.6923,
      "no_tool_cases": 4,
      "no_tool_cases_passed": 4,
      "false_positive_tool_rate": 0.0,
      "finalization_cases": 3,
      "finalization_passed": 3,
      "finalization_rate": 1.0,
      "finish_reason_tool_calls": 13,
      "content_text_tool_leaks": 0,
      "category_metrics": {
        "browser": {
          "passed": 4,
          "total": 4,
          "rate": 1.0
        },
        "calculator": {
          "passed": 2,
          "total": 2,
          "rate": 1.0
        },
        "file": {
          "passed": 2,
          "total": 3,
          "rate": 0.6667
        },
        "finalization": {
          "passed": 3,
          "total": 3,
          "rate": 1.0
        },
        "no_tool": {
          "passed": 4,
          "total": 4,
          "rate": 1.0
        },
        "terminal": {
          "passed": 1,
          "total": 4,
          "rate": 0.25
        }
      },
      "failures": {
        "generic_terminal_echo": "wrong_tool:browser_navigate",
        "generic_terminal_pwd": "wrong_tool:browser_navigate",
        "generic_terminal_ls": "wrong_tool:bash",
        "generic_search_repo": "wrong_tool:browser_navigate"
      }
    },
    "iter04_masked_colloquial_openai_parser_disabled": {
      "passed": 16,
      "total": 20,
      "overall_rate": 0.8,
      "tool_cases": 13,
      "structured_tool_cases_passed": 9,
      "structured_tool_rate": 0.6923,
      "no_tool_cases": 4,
      "no_tool_cases_passed": 4,
      "false_positive_tool_rate": 0.0,
      "finalization_cases": 3,
      "finalization_passed": 3,
      "finalization_rate": 1.0,
      "finish_reason_tool_calls": 13,
      "content_text_tool_leaks": 0,
      "category_metrics": {
        "browser": {
          "passed": 4,
          "total": 4,
          "rate": 1.0
        },
        "calculator": {
          "passed": 2,
          "total": 2,
          "rate": 1.0
        },
        "file": {
          "passed": 2,
          "total": 3,
          "rate": 0.6667
        },
        "finalization": {
          "passed": 3,
          "total": 3,
          "rate": 1.0
        },
        "no_tool": {
          "passed": 4,
          "total": 4,
          "rate": 1.0
        },
        "terminal": {
          "passed": 1,
          "total": 4,
          "rate": 0.25
        }
      },
      "failures": {
        "generic_terminal_echo": "wrong_tool:browser_navigate",
        "generic_terminal_pwd": "wrong_tool:browser_navigate",
        "generic_terminal_ls": "wrong_tool:bash",
        "generic_search_repo": "wrong_tool:browser_navigate"
      }
    },
    "current_recheck_openai_tool_calls_parser_disabled": {
      "passed": 11,
      "total": 12,
      "tool_cases": 8,
      "structured_tool_cases_passed": 8,
      "no_tool_cases": 4,
      "no_tool_cases_passed": 3,
      "finish_reason_tool_calls": 8,
      "content_text_tool_leaks": 0,
      "failures": {
        "tool_result_final_calc": "content_mismatch"
      }
    },
    "iter02_openai_tool_calls_parser_disabled": {
      "passed": 12,
      "total": 12,
      "tool_cases": 8,
      "structured_tool_cases_passed": 8,
      "no_tool_cases": 4,
      "no_tool_cases_passed": 4,
      "finish_reason_tool_calls": 8,
      "content_text_tool_leaks": 0,
      "failures": {}
    },
    "iter03_openai_tool_calls_parser_disabled": {
      "passed": 12,
      "total": 12,
      "tool_cases": 8,
      "structured_tool_cases_passed": 8,
      "no_tool_cases": 4,
      "no_tool_cases_passed": 4,
      "finish_reason_tool_calls": 8,
      "content_text_tool_leaks": 0,
      "failures": {}
    },
    "iter04_masked_openai_tool_calls_parser_disabled": {
      "passed": 12,
      "total": 12,
      "tool_cases": 8,
      "structured_tool_cases_passed": 8,
      "no_tool_cases": 4,
      "no_tool_cases_passed": 4,
      "finish_reason_tool_calls": 8,
      "content_text_tool_leaks": 0,
      "failures": {}
    }
  },
  "training_summaries": {
    "lora_tool_router_iter02": {
      "last_train": [
        "800",
        "0.068",
        "335.383",
        "92279",
        "11.108 GB"
      ],
      "last_val": [
        "800",
        "0.077"
      ],
      "peak_mem_gb": 11.108
    },
    "lora_tool_router_iter03": {
      "last_train": [
        "800",
        "0.018",
        "555.559",
        "345450",
        "11.452 GB"
      ],
      "last_val": [
        "800",
        "0.018"
      ],
      "peak_mem_gb": 11.452
    },
    "lora_tool_router_iter04_masked": {
      "last_train": [
        "600",
        "0.000",
        "25.934",
        "11599",
        "11.507 GB"
      ],
      "last_val": [
        "600",
        "0.000"
      ],
      "peak_mem_gb": 11.507
    }
  },
  "artifacts": {
    "iter02_dataset": "artifacts/repair_datasets/iter02_colloquial_router",
    "iter03_server_template_dataset": "artifacts/repair_datasets/iter03_colloquial_router_server_template",
    "iter04_chat_masked_dataset": "artifacts/repair_datasets/iter04_colloquial_router_chat_masked",
    "iter02_adapter": "artifacts/adapters/lfm_tool_router_iter02",
    "iter03_adapter": "artifacts/adapters/lfm_tool_router_iter03",
    "iter04_masked_adapter": "artifacts/adapters/lfm_tool_router_iter04_masked"
  },
  "decision": "do_not_fuse_or_publish_as_final_model_update; retain adapter artifacts and evals for analysis only"
}